Realization of a double barrier resonant tunneling diode for cavity polaritons

H. S. Nguyen1 , D.Vishnevsky2 , C. Sturm1 , D. Tanese1 , D. Solnyshkov2 , E. Galopin1 , A. Lema tre1 , I. Sagnes1 , A. Amo1 , G. Malpuech2 , and J. Bloch1
Laboratoire de Photonique et de Nanostructures, LPN/CNRS, Route de Nozay, 91460 Marcoussis , France and 2 Institut Pascal, PHOTON-N2, Clermont Universit e, Universit e Blaise Pascal, CNRS, 24 avenue des Landais, 63177 Aubi` ere Cedex, France (Dated: May 23, 2013)
1

arXiv:1305.5069v1 [cond-mat.mes-hall] 22 May 2013

We report on the realization of a double barrier resonant tunneling diode for cavity polaritons, by lateral patterning of a one-dimensional cavity. Sharp transmission resonances are demonstrated when sending a polariton ow onto the device. We use a non-resonant beam can be used as an optical gate and control the device transmission. Finally we evidence distortion of the transmission prole when going to the high density regime, signature of polariton-polariton interactions.
PACS numbers: 85.30.Mn;71.36.+c;42.65.Pc;78.55.Cr

Resonant tunneling diodes (RTD) are primary elements of nanoelectronics providing negative dierential resistance and other nonlinear properties [1, 2]. They opened the way for many applications, such as high frequency oscillation [3], resonant tunneling transistor [4] or multiple-valued logic circuits [5]. Such double barrier structures, when brought to the quantum limit, reveal fascinating physics, such as the Coulomb blockade which was observed with electrons [6] and Cooper pairs [8]. Recently, resonant transmission in high quality factor photonic crystal cavities has allowed to evidence bistable behavior and implement optical memories, making use of the carrier-induced non-linearity [7]. Transport of atomic Bose-Einstein condensates (BEC) through a double barrier structure has also been theoretically considered by several groups [9, 10], with interesting predictions related to non-linear bosonic interactions. However, no experimental demonstration of these eects for BEC has been reported so far, due to the diculty of creating adequate potential prole for atomic BEC. Cavity polaritons have appeared these last years as an alternative system to investigate the physics of out of equilibrium condensates. They are exciton-photon mixed states arising from the strong coupling between the optical mode of a cavity and excitons conned in quantum wells [11]. Cavity polaritons propagate with a speed comparable to the speed of light ( 1% speed of light) [12] thanks to their photonic component, and simultaneously show strong nonlinear interactions inherited from their excitonic component [13]. Cavity polaritons are now considered as a new platform for optical devices with many promising proposals for alloptical integrated logical circuits[1417]. Experimental demonstrations of a polariton spin switch [18], polariton transistor [19, 20] and of a polariton interferometer [21] have been recently reported . A key advantage of cavity polaritons is that the potential in which they evolve can be engineered at will, either by optical means [2224], by depositing metallic layers on top of the cavity [25], by using a surface acoustic wave [26], or by etching the cavity

into lower dimensionality microstructures [27, 28]. In this letter, we demonstrate the realization of a polariton RTD based on an innovative design of a wire cavity. Two micron-size constrictions are etched in the wire cavity and create two tunnel barriers, dening an isolated island with discrete conned polariton states. Sending a polariton ow onto this double barrier structure, we observe resonant tunneling when the polariton energy coincides with the energy of one of the conned modes. We show that a non resonant laser beam, focused onto the island, can modulate the RTD transmission with a peak to valley ratio as high as 28. The device is therefore operating as an optical gate. Finally we evidence strong asymmetry in the transmission proles when going to the high density regime. This is shown to be the signature of non-linear polariton interactions within the island, as fully supported by numerical simulations. The sample was grown by molecular beam epitaxy and consists in a /2 microcavity with 28 (resp. 40) pairs of Ga0.8 Al0.2 As/Ga0.05 Al0.95 As /4 layers in the top (resp. bottom) distributed Bragg mirror. 12 GaAs quantum wells (7 nm thickness) are inserted in the structures. The quality factor of the microcavity amounts to 100000 and the Rabi splitting to 15 meV. Electron beam lithography and inductively coupled plasma dry etching were used to fabricate 1D microwires of 3 m width and 440 m length. The RTD is dened by a microstructure described in Figs. 1(a) and 1(b): two constrictions (width = 1.4 m, length = 1 m) surround an isolated island (width = 5.6 m, length = 2 m). As the 1D connement potential is inversely proportional to the square of the wire width [29], the microstructure denes a double potential barrier and a 0D polariton island (see Fig. 1(c)). Micro-photoluminescence experiments are performed at 10K on single microwires using a cw monomode Ti:sapphire laser, focused onto a 2 m spot with a microscope objectif (NA = 0.55). A second cw monomode Ti:sapphire laser is used for experiments requiring simultaneously both resonant and non-resonant

FIG. 1. (a) Scanning electron microscopy image of the polariton RTD. (b) Width prole of the polariton RTD. (c) Simulated potential along the wire in the region of the RTD (blue), and energy of conned states within the island (red). (d) Emission spectra measured on the island for TM (red squares) and TE (blue circles) detection polarization (excitation power 300 W, laser enery 1.62 eV). (e) Spatially resolved emission of the three lowest energy conned polariton states for TM detection polarization. (f) Calculated emission pattern corresponding to (e).

FIG. 2. (a) Spectrally and spatially resolved emission measured under TM detection polarization , when exciting the structure with a non-resonant laser beam of energy 1.62 eV focused at position y=-50 m. (b)Transmission intensity measured under TM (red squares) and TE (blue circles) detection polarization. The signal is integrated between y = 7 m to y = 15 m. The red and blue lines are ts with two Lorentzians of 155 eV linewidth, limited by the detection resolution.

excitation. Polariton emission is imaged on a CCD camera coupled to a monochromator. The studied RTD corresponds to an exciton-photon detuning around -10 meV (dened as the dierence between the energy of the photonic and excitonic modes) . We rst characterize the polariton modes conned within the island, between the two tunnel barriers. Emission spectra of the island measured under TM (along yaxis) and TE (along x-axis) detection polarization are shown in Fig. 1(d). These spectra are obtained by exciting the island non-resonantly with excitation power well below condensation threshold. Three discrete polariton modes are observed for each polarization. Polarization splitting reect the anisotropy of the connement within the island. Spatial mapping of these conned modes is presented in Fig. 1(e) for TM polarization. Characteristic emission pattern with well dened emission lobes are observed in good agreement with simulations obtained solving a 2D Schr odinger equation for a particle of eective mass m = 6.1 105 melectron in a potential corresponding to that of the heterostructure. These measurements demonstrate that the two constrictions dene an isolated island with well dened discrete states. Let us now discuss polariton transport through this

double barrier structure. We rst send onto the microstructure a polariton ow with a broad energy distribution. For this experiment, polaritons are injected with non resonant excitation far from the microstructures (typically 50 to 80 m away). Figure 2(a) displays the spectral emission measured along the wire in TM detection polarization. The transmitted signal (y > 3 m) presents a sharp threshold energy above which polaritons propagate across the island (E > Eb = 1575.8 meV). This energy corresponds to the top of the tunnel barrier, and we deduce a barrier height of Vb 3.0 meV. Below Eb , the transmission is vanishing except for a few sharp resonances. The lowest one is observed at 1573.7 meV, and corresponds to the resonance of incident polaritons with the 1T M mode of the island. Transmission spectra for energy close to this resonance are reported in Fig. 2(b) for both detection polarizations: the observed peaks are splitted by the TE-TM splitting, showing that polariton polarization is preserved in the resonant tunneling. Notice that resonant tunneling transmission corresponding to 2T M (i.e. 1575.0 meV) is weak in Fig. 2(a). This feature can be explained considering the mode mismatch between the 1D incident polariton mode (symmetric along x) and the 2T M mode (antisymmetric along x). Finally, notice that the fringe pattern observed in the upstream side (y < 3 m) of Fig. 2(a) is due to the interferences between incident polaritons and polaritons reected on the double barrier structure. In the following, we describe how the transmitted intensity can be optically modulated, using a second laser beam focused on the island. To do this, we use the repul-

3 which corresponds to an induced blueshift of 111 eV. On the opposite, vanishing transmitted signal is observed for a blueshift of 25 eV or 300 eV. These results prove that our device is indeed operated as an all-opticallycontrolled RTD, with a very high spectral selectivity. The peak (valley) transmission amounts to 11.1 % (0.4%) resulting in a Peak-To-Valley Signal Ratio (PVSR) of 28 [30]. An interesting feature is revealed in the transmission spectrum presented in Fig. 3(b): the line-shape is distorted with respect to a Lorentzian prole, with a more abrupt shape on the low power side. We show below that this is a direct evidence of polariton-polariton interactions [13] within the island, resulting in a nonlinear tunneling regime of the device. Let us rst give a qualitative explanation of the observed asymmetric transmission prole. When PGate is slightly lower than the resonance power Pres , incident polaritons start to enter the island. As a consequence, the energy of 1T M undergoes an additional blueshift due to the interactions between polaritons conned in the island, which reduces the detuning between EInc and the energy of 1T M . This mechanism thus provides a positive feedback that accelerates the passage to the resonant tunneling. On the contrary, when PGate slightly exceeds Pres , less polaritons enter the island. Thus, the additional blueshift decreases and the passage to o-resonant tunneling is slowed down due to a negative feedback. Of course this non-linear regime occurs when the polariton density of the incident ow is large enough. If we reduce PInc suciently to enter the linear regime, then the luminescence from the island induced by PGate starts to dominate the emission spectra, and precise extraction of the transmission proles becomes delicate. To unambiguously demonstrate the two regimes (linear and nonlinear), we performed experiments with a single laser beam (the one injecting the polariton ow) and probe the transmission prole when scanning the incident energy EInc . For low incident power, a symmetrical transmission spectrum is measured characteristic of the linear regime (see Figure 4(a) corresponding to PInc = 5 mW ). The transmission prole is well tted by a Lorentzian of linewidth = 27 eV , attributed to the 1T M mode homogeneous linewidth. In analogy to the non dissipative case, where the tunneling transit time transit [31] is twice the particle lifetime in the island[32], we can write here: = 2 /transit + /rad , where rad is the polariton radiative lifetime within the island, governed by the escape of the photon through the mirrors or the sidewalls the structure. Moreover the value of the transmission peak is given by Tres = rad /(rad + transit ) 0.165. Thus we deduce transit = 172 ps and rad = 34 ps. Notice that rad is close to the nominal photon lifetime calculated from the quality factor of the non-etched structure, expected to be around 40 ps. For larger PInc , asymmetry of the transmission spectrum develops as illustrated in

FIG. 3. (a) (black circles) Measured blueshift of the 1T M mode as function of the optical gate power PGate . (red line) Lineart with a 174 eV.mW 1 slope. The energy of the laser gate beam is 1.62 eV . (b) Tunneling transmission as a function of the blueshift of the 1T M mode (bottom axis), or of PGate (top axis). The red line is a guide to the eye. The polariton ow is injected at y = 80 m by a resonant laser of power PInc = 40 mW and of energy EInc = 1573.8 meV . (ce) Spatially resolved emission measured for dierent values of PGate corresponding to a blueshift of the 1T M mode equal to (c) 25 eV , (d) 111 eV and (e) 300 eV . (f-h) Corresponding measured integrated intensity as a function of position (blue) and wire potential (red).

sive interaction of polaritons with a reservoir of excitons [23], which is locally injected in the island using a weak non-resonant optical excitation of power PGate . When PGate is turned on, polariton-exciton interactions induce a 174 eV.mW 1 blueshift of the polariton modes conned within the island (see Figure 3(a)). We use this gate beam to control the transport of a monochromatic TM polarized polariton ow, which is sent onto the double barrier structure, using a resonant laser beam (of energy EInc and power PInc ). We chose EInc = 1573.8 meV , a value slightly larger than the energy of the 1T M mode (which lies at 1573.7 meV). Figure 3(b) reports the measured transmission for dierent values of PGate . When the gate beam brings the energy of the 1T M mode into resonance with the polariton ow, a pronounced increase in the transmission is induced. This is further illustrated in Fig. 3(c-h), where spatially resolved emission is monitored for values of PGate corresponding to energy of the 1T M mode below, at and above the resonance. We clearly observed an enhanced luminescence signal in the downstream region at resonance (see Fig. 3(d,g)),

4 Fig 4(b) for PInc = 40 mW . This asymmetrical prole, induced by polariton-polariton interactions, is a mirror image of the one shown in Fig. 3(b), as here EInc is scanned with respect to the energy of the 1T M mode, whereas before the energy of 1T M was tuned with respect to EInc using the gate beam. Such asymmetric shape of transmission prole due to nonlinear interaction has already been predicted for cold atom condensate transport through a double potential barrier [10], and also been observed for resonant tunneling of guided light propagation through a photonic crystal microcavity [7]. To describe our experiments, we developed numerical simulations taking into account not only polariton-polariton interactions, but also the nite polariton lifetime, which is the specicity of our dissipative system. Figure. 4(c) presents the numerical simulations of transmission prole in the linear and nonlinear regime obtained with an interaction constant g = 0.4 eV.m and a polariton density of 65 m1 and 390 m1 . A 1D Gross-Pitaevskii equation has been used for Fig. 4(c), and the validity of this approximation has been conrmed by 2D simulations. The calculated proles reproduce the observed features. However, while the simulation shown in Fig. 4(c) and the theoretical calculation for atom condensates [10] predict a blueshift of the resonant tunneling peak in the nonlinear regime with respect to the one in the linear regime, a redshift is observed in our experimental results (see Figs. 4(a) and . 4(b)). We think that this discrepancy between theory and experiment is due local heating of the sample in the high density regime. Finally we would like to discuss the performance of our polariton RTD device. In the present experiment, the peak transmission is limited to 16.5% because of the long tunneling time as compared to the photon lifetime. Further engineering of the tunnel barrier could allow controlling the tunneling transit time and thus optimizing the transmission. Indeed, Fig. 4(d) presents simulation of the resonant transmission and the tunneling transit time corresponding to dierent barrier heights. Our simulation points that resonant transmission coecient can be signicantly increased when reducing the barrier height. Concerning the dynamics of our RTD device, the optical gating dynamic is expected to be limited by the lifetime of the excitonic reservoir, which is around 400 ps [33]. Thus, we estimate that our device could operate at frequency of several GHz. We could envisage to increase this operating speed by using a resonant optical gate. Indeed then the dynamics would be mainly governed by the polariton lifetime which is much shorter. In conclusion, we have fabricated a polariton RTD exhibiting resonant polariton transport through a double potential barrier structure. The device is gated by a lowpower non-resonant laser, which modulates the transmission within a PVSR of 28. Nonlinear transmission regime of a BEC is demonstrated experimentally for the rst time. These results open the way for a new generation

FIG. 4. (a,b) Tunneling transmission measured as a function of EInc for (a) PInc = 5 mW and (b) PInc = 40 mW . The red line in (a) is a Lorentzian t of linewidth 27 eV , while the red line in (b) is a guide to the eye. The polariton ow is injected at y = 80 m. (c) Simulation of the resonant tunneling transmission prole for low pump density (red circles) and high pump density (blue squares). Simulation parameters: polariton radiative lifetime rad = 34 ps, barrier potential Vb = 2.7 meV , interaction constant g = 0.4 eV.m, incident polariton density 65 m1 (resp. 390 m1 ) for the low (resp. high) pump density case. (d) Simulation of the resonant transmission Tres (red circles) and the tunneling transit time transit (blue squares) as a function of the barrier height.

of integrated logical circuit by exploiting for instance the nonlinear transport of many RTDs in a more complex architecture [17]. Moreover, by reducing the size of the isolated island, the quantum regime could be reached: resonant tunneling transmission could become sensitive to single-polariton non-linearity [34]. In this regime, emission of non classical light is expected because of polariton blockade, together with many fascinating features of the Bose-Hubbard physics [3538]. We thank Luc Le Gratiet for scanning electron microscopy. The work was supported nancially by the Agence Nationale pour la Recherche project Perocai and Quandyde (ANR-11-BS10-001), by the FP7 ITN Clermont4 (235114) and Spin-Optronics (237252), by the french RENATECH network and the RTRA Triangle de la Physique contract Boseow1D and 20120397-Interpol.

5
[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] T. Gao et al., Phys. Rev. B 85, 235102 (2012). D. Ballarini et al., arXiv:1201.4071. C. Sturm et al., arXiv:1303.1649. E. Wertz et al., Nat. Phys. 6, 860 (2010). A. Amo et al., Phys. Rev. B 82, 081301 (2010). G. Tosi et al., Nat. Phys. 8, 190 (2012). C. W. Lai et al., Nature 450, 529 (2007). E. A. Cerda-M endez et al., Phys. Rev. Lett. 105, 116402 (2010). J. Bloch et al., Superlattices Microstruct. 22, 371 (1997). T. K. Para so et al., Phys. Rev. B 79, 045319 (2009). G. Dasbach et al., Phys. Rev. B. 66, 201201 (2002). See Supplemental Material for a movie showing the measured emission along the wire for diferent values of the optical gate power. M. B uttiker and R. Landauer, Phys. Rev. Lett. 49, 1739 (1982). U. Peskin et al., J. Phys. Chem. B 106, 33 (2002). D. Bajoni et al., Phys. Rev. B 73, 205344 (2006). A. Verger et al., Phys. Rev. B 73, 193306 (2006). T. C. H. Liew and V. Savona, Phys. Rev. Lett. 104, 183601 (2010). M. Bamba, Phys. Rev. A 83, 021802 (2011). A. D. Greentree et al., Nat. Phys. 2, 856 (2006). M. J. Hartmann et al., Nat. Phys. 2, 849 (2006).

jacqueline.bloch@lpn.cnrs.fr [1] T. C. L. Sollner et al., Appl. Phys. Lett. 43, 588 (1983). [2] H. Mizuta and T. Tanoue, The physics and applications of resonant tunneling diodes (Cambridge University Press, 1995). [3] E. R. Brown et al., Appl. Phys. Lett. 58, 2291 (1991). [4] F. Capasso and R. Kiehl, J. Appl. Phys. 58, 1366 (1985). [5] P. Mazumder et al., Proceedings of IEEE. 86, 664 (1998). [6] C. Pasquier et al., Phys. Rev. Lett. 70, 69 (1993). [7] K. Nozaki et al., Nat. Phot. 6, 248 (2012). [8] P. Joyez et al., Phys. Rev. Lett. 72, 2458 (1994). [9] I. Carusotto et al., Phys. Rev. Lett. 84, 399 (2000). [10] T. Paul et al., Phys. Rev. Lett. 94, 020404 (2005). [11] C. Weisbuch et al., Phys. Rev. Lett. 64, 3314 (1992). [12] T. Freixanet et al., Phys. Rev. B 61, 7233 (2000). [13] C. Ciuti et al., Phys. Rev. B 63, 041303(R) (2001). [14] T. C. H. Liew et al., Phys. Rev. Lett. 101, 016402 (2008). [15] I. A. Shelykh et al., Phys. Rev. Lett. 102, 046407 (2009). [16] I. A. Shelykh et al., Phys. Rev. B 82, 153303 (2010). [17] T. Espinosa-Ortega and T. C. H. Liew, arXiv:1302.1935. [18] A. Amo et al., Nat. Phot. 4, 361 (2010).

[31] [32] [33] [34] [35] [36] [37] [38]